Hello,
I have an important new information to report that may make the bisected commit not (entirely) guilty. It may be more of a sign of a subtle hardware failure on my end than a kernel regression/bug.
The fact that the bisection I did gave out the very first commit after the v5.12.8 tag puzzled me. So I built the 5.12.8 version with the same `.config` (I attached to this mail as `config.bisect`) that I used for bisection and worked with it since then, it took hours (instead of the average dozen minutes) but it ended up freezing. So the changes that have been brought by the ALSA patches that start with the commit 73d862c55 seem to have only made a deeper, more subtle, issue more apparent by making the freeze happen very early.
The 5.12.8 version that I use that hasn't frozen for the past weeks of use contains out-of-tree patches (e.g. a different CPU task scheduler) and does not have the same `.config` file (I attached it as `config.works`). `config.works` freezes with 5.12.9 too (with the same extra out-of-tree patches). I will reuse it as a base for vanilla building and test things out again. I will also reduce my RAM frequency to a very safe (not XMP) one.
I am willing to do everything that is needed to pinpoint precisely the issue on my own, with your help, if possible.
I hope I am not wasting anyone's time with this report. If you deem it so, please inform me and I will immediately refrain from writing any further.
Kind regards,
Adel K.S.